provide multiple ways of training genemark and using hints for prediction #24

Author: Jon Palmer

CITATION.cff

@@ -1,4 +1,4 @@
-cff-version: version = "25.7.20"
+cff-version: version = "25.8.3"
 title: 'funannotate2: eukaryotic genome annotation'
 message: >-
   If you use this software, please cite it using the
@@ -17,5 +17,5 @@ keywords:
   - functional annotation
   - consensus gene models
 license: BSD-2-Clause
-version: version = "25.7.20"
-date-released: '2025-07-20'
+version: version = "25.8.3"
+date-released: '2025-08-04'

MEMORY_MONITORING.md

@@ -54,27 +54,30 @@ The memory monitoring is integrated into:
 
 ### Command Line Options
 
-Add memory monitoring to funannotate2 predict:
+Memory monitoring is enabled by default in funannotate2 predict:
 
 ```bash
-# Enable memory monitoring
-funannotate2 predict -i input_dir --monitor-memory
+# Memory monitoring is enabled by default
+funannotate2 predict -i input_dir
 
-# Enable memory monitoring with memory limit
-funannotate2 predict -i input_dir --monitor-memory --memory-limit 16
+# Set memory limit with memory monitoring (default behavior)
+funannotate2 predict -i input_dir --memory-limit 16
+
+# Disable memory monitoring if needed
+funannotate2 predict -i input_dir --disable-memory-monitoring
 ```
 
 ### CLI Options
 
-- `--monitor-memory`: Enable memory monitoring and prediction
+- `--disable-memory-monitoring`: Disable memory monitoring and prediction (enabled by default)
 - `--memory-limit GB`: Set memory limit in GB to adjust CPU allocation
 
 ### Example Output
 
-When memory monitoring is enabled, you'll see output like:
+With memory monitoring enabled by default, you'll see output like:
 
 ```
-Memory monitoring enabled for ab initio predictions
+Memory monitoring: using 14.4 GB limit (90% of 16.0 GB total)
 Memory limit set to 16.0 GB
 Memory usage estimate for 150 contigs with tools ['snap', 'augustus']:
   Total estimated peak memory: 2847.3 MB

funannotate2/__main__.py

@@ -276,9 +276,9 @@ def predict_subparser(subparsers):
         "--tmpdir", default="/tmp", help="volume to write tmp files", metavar=""
     )
     optional_args.add_argument(
-        "--monitor-memory",
+        "--disable-memory-monitoring",
         action="store_true",
-        help="Monitor memory usage of ab initio prediction tools",
+        help="Disable memory monitoring of ab initio prediction tools (enabled by default)",
     )
     optional_args.add_argument(
         "--memory-limit",
@@ -372,6 +372,13 @@ def train_subparser(subparsers):
         help="Maximum number of gene models to use for training (default: 5000)",
         metavar="",
     )
+    optional_args.add_argument(
+        "--genemark-mode",
+        default="fast",
+        choices=["fast", "unsupervised", "guided"],
+        help="GeneMark training mode: fast (subset+hints, default), unsupervised (self-training), guided (full+hints)",
+        metavar="",
+    )
     optional_args.add_argument(
         "--busco-lineage",
         dest="busco_lineage",

funannotate2/train.py

@@ -348,18 +348,20 @@ def train(args):
 
     # check if genemark is installed, if so then run self training
     if which_path("gmes_petap.pl") and which_path("gmhmme3"):
-        logger.info("Training GeneMark-ES using self-training")
+        logger.info(f"Training GeneMark-ES using {args.genemark_mode} mode")
         fungus_flag = False
         if taxonomy["kingdom"] == "Fungi":
             fungus_flag = True
         genemark_train = train_genemark(
             TrainingGenomeFasta,
-            train_models,
+            filt_train_models_final,
             test_models,
             folder=misc_dir,
             fungus=fungus_flag,
             cpus=args.cpus,
             log=logger,
+            training_mode=args.genemark_mode,  # User-configurable training mode
+            max_training_length=50000000,  # 50 Mb limit for additional genomic context
         )
         genemark_train["training_set"] = "self training"
         train_data["genemark"] = genemark_train

funannotate2/utilities.py

@@ -833,16 +833,18 @@ def logoutput(logname, process_result):
 
                 # Try to determine the output directory from environment or current working directory
                 output_dir = os.environ.get("FUNANNOTATE2_OUTPUT_DIR", os.getcwd())
-                logs_dir = os.path.join(output_dir, "logs")
+                logs_dir = os.path.join(output_dir, "logfiles")
 
-                # Create logs directory if it doesn't exist
+                # Create logfiles directory if it doesn't exist
                 if not os.path.exists(logs_dir):
                     try:
                         os.makedirs(logs_dir)
                     except:
                         logs_dir = output_dir  # Fallback to output directory
 
-                memory_log_file = os.path.join(logs_dir, "memory-monitoring.jsonl")
+                memory_log_file = os.path.join(
+                    logs_dir, "predict-abinitio-memory-monitoring.jsonl"
+                )
 
                 # Extract additional metadata for prediction model improvement
                 tool_name = None

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "funannotate2"
-version = "25.7.20"
+version = "25.8.3"
 description = "Funannotate2: eukarytoic genome annotation pipeline"
 readme = {file = "README.md", content-type = "text/markdown"}
 authors = [